<<=> not handled by extension "mhchem"

mikkelricky commented 12 years ago

The reaction arrow <<=> is parsed as < followed by <=> , but should be parsed as a single arrow and rendered as "<=>> rotated 180 degrees" (I'm not very strong at chemistry).

I've tried to fix this myself and have come up with the following (expressed as a git diff)

diff --git a/unpacked/extensions/TeX/mhchem.js b/unpacked/extensions/TeX/mhchem.js
index 5326885..704f63c 100644
--- a/unpacked/extensions/TeX/mhchem.js
+++ b/unpacked/extensions/TeX/mhchem.js
@@ -81,6 +81,7 @@ MathJax.Hub.Register.StartupHook("TeX Jax Ready",function () {
       '<->': "leftrightarrow",
       '<=>': "rightleftharpoons",
       '<=>>': "Rightleftharpoons",
+      '<<=>': "Leftrightharpoons",
       '^': "uparrow",
       'v': "downarrow"
     },
@@ -204,7 +205,7 @@ MathJax.Hub.Register.StartupHook("TeX Jax Ready",function () {
     //
     ParseLess: function (c) {
       this.FinishAtom();
-      var arrow = this.Match(/^(<->?|<=>>?)/);
+      var arrow = this.Match(/^(<->?|<?<=>>?)/);
       if (!arrow) {this.tex += c; this.i++} else {this.AddArrow(arrow)}
     },

@@ -374,10 +375,12 @@ MathJax.Hub.Register.StartupHook("TeX Jax Ready",function () {
   MACROS.xleftrightarrow = ['xArrow',0x2194,6,6];
   MACROS.xrightleftharpoons = ['xArrow',0x21CC,5,7];  // FIXME:  doesn't stretch in HTML-CSS output
   MACROS.xRightleftharpoons = ['xArrow',0x21CC,5,7];  // FIXME:  how should this be handled?
+  MACROS.xLeftrightharpoons = ['xArrow',0x21CC,5,7];

   //  FIXME:  These don't work well in FF NativeMML mode
   MACROS.longrightleftharpoons = ["Macro","\\stackrel{\\textstyle{{-}\\!\\!{\\rightharpoonup}}}{\\smash{{\\leftharpoondown}\\!\\!{-}}}"];
   MACROS.longRightleftharpoons = ["Macro","\\stackrel{\\textstyle{-}\\!\\!{\\rightharpoonup}}{\\small\\smash\\leftharpoondown}"];
+  MACROS.longLeftrightharpoons = ["Macro","\\stackrel{\\rightharpoonup}{{{\\leftharpoondown}\\!\\!\\textstyle{-}}}"];

   //
   //  Needed for \bond for the ~ forms

dpvc commented 12 years ago

Thanks for the information. The <<=> arrow doesn't seem to be in the documentation that I have, which is why I didn't include it. Your changes look good except for the match pattern, which would also allow <<=>>, so I would use /^(<->?|<==>?|<<=>)/ instead.

I will make the modifications to the mhchem package.

mikkelricky commented 12 years ago

Cool. And you're absolutely right that my match pattern isn't as precise as it should be. The correct(ed) pattern in your comment contains a minor typo, but I suppose we can agree that the pattern should be /^(<->?|<=>>?|<<=>)/.

dpvc commented 12 years ago

Right. Your final pattern is the right one. I'm reopening the issue so that I will be sure to make the changes, and will close it again after that has occurred.

fred-wang commented 12 years ago

I started to write tests for the mhchem extension: https://github.com/fred-wang/MathJax-test/tree/master/testsuite/LaTeXToMathML/mhchem

These are based on example from http://mirror.ctan.org/macros/latex/contrib/mhchem/mhchem.pdf

I added a test for an unknown bond and for the "<<=>" arrow mentioned in this issue.

I have a [Math Processing Error] with advanced-1.html. For various other tests I get undefined macros. It seems that adding the mhchem extension does not force the loading of the package containing extended arrow definitions.

Maybe some tests are not correct or I did not use the correct MathJax version.

dpvc commented 12 years ago

Right, the mhchem extension relies on the AMSmath extension, but it doesn't currently force it to be loaded. I will have to do that. In the meantime, if you add "AMSmath.js" to the TeX.extensions array, that should make the tests work.

Davide

fred-wang commented 12 years ago

OK, I've updated the tests on my fork. I still have a [Math Processing Error] with advanced-1.html. It seems that the "<=>" arrow does not work in that case (it works in arrows-1.html, though).

dpvc commented 11 years ago

The issue234 branch of my fork of MathJax should resolve both the missing <<=> arrow, and the [Math Processing Error] that Fred reported. The mhchem extension now sets things up to load AMSmath automatically when needed and properly synchronizes with its loading. I think the processing error came from not synchronizing properly so that mhchem tried to use part of AMSmath before it was loaded. The is certainly the issue with the undefined macros. In any case, the test files seem to work for me now. (And thanks for putting them together.)